A portfolio approach to improving market and credit risk management

The credit crisis (which began in August 2008) has affected almost every segment of the international financial system. Credit has been severely curtailed as banks struggle to contain further losses caused by reckless lending practices that characterised the last two decades. Asset prices have tumbled as fearful investors flee to safer havens, abandoning traditional investments and hedge funds with reso- lute consistency. Governments – in an attempt to stave off stagflation and kick-start failing economies – have reduced interest rates to historic lows, initiated stimulus packages and instigated bank bailouts, but the efforts have (as yet) had minimal to no effect on markets. The dire economic environment characterised by diminishing industrial production, falling house (and other asset) prices and rising unemployment, has only discouraged spending and investing and promoted capital hoarding. In the ensuing crisis, the regulatory economic environment (dominated by the Basel Committee for Banking Supervision's (BCBS) Basel II Accord) has proved inadequate. Potential solutions have not yet pre- sented themselves and the crisis looks likely to continue for the foreseeable future. In the light of these events and contemporary failings of finance in general, the need to continuously augment existing and invent new techniques to measure and manage financial risks are paramount. This thesis explores four significant problems facing modern risk management in a portfolio context. The first problem examines the assumption of normally distributed portfolio returns. Compelling evidence for the consistent failure of this assumption is provided. A measure for ranking portfolio per- formance is discussed and explained with reference to several South African hedge fund portfolios. The second of these problems explores the assumption of unlimited liquidity in market risk measurement models. This assumption has been shown to be utterly fallacious and indeed, is now be- lieved to be the principal component of the credit crisis. A new portfolio market risk model, which incorporates the effect of severely diminished liquidity, is introduced and applied to several South African market portfolios. The results indicate a substantially improved model of market risk. The third problem probes the effect of obligor default quality discrimination to address a subtle discrepancy in the BCBS's formulation for credit portfolio capital charges. The cause of this discrepancy is located and its effects discussed with far reaching consequences for retail loan portfolios. Finally, the lack of a robust technique to extract retail asset correlations from empirical loan loss data is investigated. A methodology is devised using the underlying BCBS formulation for credit risk and the results obtained are compared with retail asset correlations stipulated by the BCBS. The empirical correlations (and the associated capital charges) were found to be considerably lower than the BCBS correlations (and capital charges), even during the elevated losses currently (2009) being experienced. The accuracy of these punitive impositions in a portfolio context is assessed and suggestions are made for further empirical study.